import torch
import streamlit as st
from transformers.models.bart import BartForConditionalGeneration
from transformers import PreTrainedTokenizerFast

#@st.cache
#@st.cache_data(allow_output_mutation=True)
@st.cache_data()
def load_model():
    #model = BartForConditionalGeneration.from_pretrained('logs/model_chp/epoch-6')
    model = BartForConditionalGeneration.from_pretrained('LeeJang/news-summarization-v2')
    # tokenizer = get_kobart_tokenizer()
    return model

model = load_model()
tokenizer = PreTrainedTokenizerFast.from_pretrained('gogamza/kobart-base-v1')
st.title("2문장 뉴스 요약기")
text = st.text_area("뉴스 입력:")

st.markdown("## 뉴스 원문")
st.write(text)

#'''
if text:
    text = text.replace('\n', ' ')
    text = text.strip()
    
    arr = text.split(' ')

    if len(arr) > 501:
        #print('!!!')
        arr = arr[:501]
        text = ' '.join(arr)
    
    st.markdown("## 요약 결과")
    with st.spinner('processing..'):
        input_ids = tokenizer.encode(text)
        input_ids = torch.tensor(input_ids)
        input_ids = input_ids.unsqueeze(0)
        output = model.generate(input_ids, eos_token_id=1, max_length=512, num_beams=5)
        output = tokenizer.decode(output[0], skip_special_tokens=True)
    st.write(output)
#'''